https://repository.hneu.edu.ua/handle/123456789/139 2026-07-20T11:10:58Z 2026-07-20T11:10:58Z Знахур С. https://repository.hneu.edu.ua/handle/123456789/41444 2026-07-20T09:17:42Z 2026-01-01T00:00:00Z

Назва: Neural network optimisation: quantisation techniques, practical benchmarking, and deployment recommendations Автори: Знахур С. Короткий огляд (реферат): The paper addresses the deployment of deep convolutional neural networks under the memory and energy limitations typical of mobile hardware, focusing on quantisation as a practical optimisation technique. The author performs a systematic empirical comparison of five convolutional architectures (MobileNetV2, ResNet-50, DenseNet-121, EfficientNet-B3, EfficientNet-B4), each converted with TensorFlow Lite Converter into three numerical-precision formats — FP32 baseline, FP16 and INT8 — and benchmarked on a physical iPhone 13 device (Apple A15 Bionic) through a purpose-built React Native 0.82 application using the react-native-fast-tflite library. Classification accuracy is assessed on a 500-image ImageNet-v2 subset. The results demonstrate that FP16 conversion halves model size while keeping Top-1 accuracy loss below 0.3% for every architecture, whereas INT8 reduces file size by 71–74% and improves inference latency by up to 2.8 times for mobile-friendly networks, but yields no consistent latency gain for EfficientNet-B3/B4 because part of their computation graph falls back to CPU execution within the TensorFlow Lite Metal delegate. Three factors governing quantisation effectiveness are identified: the structural compatibility of an architecture with the hardware delegate, the representativeness of the calibration dataset, and the delegate's operator coverage. Based on these findings, the author formulates differentiated, evidence-based deployment recommendations for mobile AI developers, organised by architecture family and use-case requirements, and proposes a reusable conversion-and-benchmarking methodology applicable beyond the five architectures studied.

2026-01-01T00:00:00Z Знахур Л. https://repository.hneu.edu.ua/handle/123456789/41441 2026-07-20T09:12:10Z 2026-01-01T00:00:00Z

Назва: Mobile information system based on optimized deep neural networks for resource-constrained platforms Автори: Знахур Л. Короткий огляд (реферат): The paper is devoted to designing a mobile information system built around optimised deep neural networks intended for resource-constrained platforms. The author provides a comprehensive analysis of model-optimisation techniques — numerical quantisation, structural pruning and knowledge distillation — with a primary focus on post-training quantisation (PTQ) and quantisation-aware training (QAT), and formulates the underlying scale-factor and zero-point mapping equations used for INT8 conversion. The practical part experimentally evaluates five convolutional architectures — MobileNetV2, ResNet-50, DenseNet-121, EfficientNet-B3, EfficientNet-B4 — converted with the TensorFlow Lite framework and benchmarked with a single-threaded React Native CLI application on a physical Apple iPhone (A14 Bionic). Four metrics are analysed for every model and precision format: disk footprint, inference latency, classification accuracy, and resource efficiency. The results confirm that FP16 conversion halves storage requirements with a negligible accuracy loss of up to 0.1%, while full INT8 static quantisation reduces file size by up to 75% and accelerates inference by up to 62.1%, at the cost of accuracy degradation ranging from 0.6% for MobileNetV2 to 2.4% for EfficientNet-B4. This variation is linked to the structural properties of each network: depthwise separable convolutions and linear bottlenecks in MobileNetV2, together with the residual shortcuts of ResNet-50, provide strong resilience to fixed-point rounding errors, whereas the dense feature concatenation of DenseNet-121 and the tightly balanced compound scaling of the EfficientNet family make them considerably more sensitive to INT8 conversion. Part of the observed latency behaviour is attributed to the fallback of unsupported operators from INT8 kernels to FP32 CPU execution inside the TensorFlow Lite interpreter. Qualitative thermal observations, supported by published hardware benchmarks, additionally indicate 35–50% lower energy consumption for INT8 models compared with FP32 baselines. The findings support recommending quantisation-aware training for architecture families most sensitive to static INT8 quantisation, so as to preserve classification accuracy while retaining the compression and efficiency benefits required for mobile deployment.

2026-01-01T00:00:00Z Знахур С. https://repository.hneu.edu.ua/handle/123456789/41430 2026-07-17T10:21:04Z 2026-01-01T00:00:00Z

Назва: Evaluation frameworks for large language models Автори: Знахур С. Короткий огляд (реферат): The paper addresses the challenge of assessing quality in large language model (LLM)-based systems, whose stochastic, non-deterministic outputs make classical pass/fail testing insufficient. The author proposes and empirically validates a structured, multi-method quality-assurance (QA) framework that combines four complementary evaluation strategies — an LLM-as-a-Judge holistic rubric, GEval reference-based chain-of-thought scoring, AspectCritic binary aspect-level verification, and SelfCheckBERTScore semantic-consistency measurement across repeated samples. The framework is applied to a locally hosted LLL relation model run through the Ollama runtime and evaluated on a purpose-built dataset of 500 prompts spanning factual, relational, and open-ended analytical categories. Pass rates across the four metrics range from 62% (GEval) to 76% (SelfCheckBERTScore), with performance declining consistently from factual to relational to analytical prompts — a gradient the author attributes to measurable prompt-level features such as length, number of referenced entities, and the presence of contrastive or negation cues. Pairwise inter-metric agreement, computed with Cohen's Kappa, shows substantial agreement between content-oriented metrics (LLM-as-a-Judge vs. AspectCritic, κ = 0.61; AspectCritic vs. GEval, κ = 0.65) but only fair agreement between SelfCheckBERTScore and the correctness-oriented metrics (κ = 0.35–0.43), confirming that output consistency and content correctness are structurally distinct quality dimensions. Manual failure analysis identifies three primary failure patterns — partial incompleteness (46% of failures), factual imprecision (32%), and semantic inconsistency (21%) — each detected preferentially by a different subset of metrics. Based on these findings, the author proposes three targeted, low-cost improvements: structured prompt templates that force explicit enumeration of relational components, GEval-based post-generation filtering as an internal quality gate, and reduced sampling temperature for analytical prompts to curb cross-sample divergence. The study demonstrates that no single evaluation method captures the full spectrum of LLM quality concerns and contributes a reproducible, API-independent evaluation pipeline applicable to other LLM-based systems.

2026-01-01T00:00:00Z Знахур Л. https://repository.hneu.edu.ua/handle/123456789/41429 2026-07-17T10:15:25Z 2026-01-01T00:00:00Z

Назва: AI-driven information systems development: LLM-based code generation approaches Автори: Знахур Л. Короткий огляд (реферат): The paper presents a systematic review and empirical investigation of large language model (LLM)-based code generation systems covering the period 2022–2026, together with a proposed hybrid architecture that combines Retrieval-Augmented Generation (RAG) with a Mixture-of-Experts (MoE) routing strategy. The author surveys the three principal model families used for code generation — autoregressive decoders, masked language models, and encoder-decoder architectures — and synthesises benchmark results for leading systems such as Claude 3.7 Sonnet, GPT-4o, Gemini 2.5 Pro, DeepSeek V3 and Llama 4 Maverick across HumanEval, SWE-Bench, MBPP and LiveCodeBench. A Python-based prototype of the proposed hybrid model is implemented using QZhou-Embedding for query representation, a FAISS vector index for retrieval over a 10,000-snippet Python code corpus, three specialised MoE expert modules (Python, JavaScript, and security analysis), and an iterative self-reflection loop that verifies generated code through AST parsing and chain-of-thought semantic checks. Evaluated on the full 164-problem HumanEval benchmark, the hybrid model achieves a Pass@1 of 85%, exceeding CodeLlama by 15 percentage points and GPT-4o by 11 percentage points, while also producing code with a higher pylint score (8.5/10), lower cyclomatic complexity (2.3), and a higher Maintainability Index (85) than both baselines. The retrieval subsystem is shown to add less than 1% to end-to-end query latency while the generation stage accounts for roughly 94%, and the system scales acceptably under concurrent load. The author identifies three recurring limitation patterns — retrieval noise, self-reflection runtime overhead, and limited support for low-resource programming languages — and proposes fine-grained candidate re-ranking, adaptive reflection termination, and multilingual corpus extension as mitigation directions. The findings support the conclusion that combining retrieval grounding, task-specialised routing, and iterative verification yields measurable improvements in functional correctness and code quality over both commercial and open-source baselines.

2026-01-01T00:00:00Z